Step 4 – Extraction approach
Based on the requirements for extraction and volume of documents, the approach contract extraction is determined - whether manual or using an AI/ML solution. Manual extraction is the best suited option when the volume of contracts to be extracted is low; or the type of attributes to be extracted is complex or subject to interpretation.
An AI/ML solution can be leveraged for high volume/low complexity extraction. The AI solution can significantly help in reducing the time and effort required for attribute extraction.
However, it is important to understand the limitation of AI. The AI solutions offer standard attribute extraction features that enable search and reporting capabilities. To enable key insights into deliverables, service levels and contractual risks it is often required to train the AI solution to be able to extract the output. In addition,, it is important to build a process that provides validation of the AI output to ensure that the extracted data meets the quality standards defined by client.
Wipro’s ‘AI with lawyer in the loop model’ combines the legal expertise and AI/ML capabilities to deliver faster and quality contract extraction. Our team of legal experts support in training of the AI platform as well as validation of tool output before the extracted data is uploaded into the CLM platform.
Step 5 – Training and AI configuration
If the client opts for the AI solution for contract extraction and depending on the complexity of the requirement, the AI/ML solution needs to be trained on client specific contract types and configured to address the use cases defined as part of the legacy migration process.
The approach for ML model training and testing is as follows:
- Segregate documents and identify variations and samples for labeling
- Identify the approach for tagging of attributes
- Label sample documents depending on the complexity and variation to train the AI model
- Train the ML model and test based on the labelled documents. Conduct additional labelling and model training based on the test results
- Revise the ML model based on additional variations observed in the larger data
- Validate tool output to ensure quality of deliverables
Step 6 – Ongoing extraction
Once the AI/ML tool is trained, the team proceeds with ongoing extraction and validation of documents in batches.
A large number of organizations are adopting a CLM platform as a solution for efficient management of their contract portfolio. However, majority of them ignore the relevance of migrating legacy contracts to the platform to unleash the true value of the implemented solution.
A well-defined contract extraction and legacy migration process can help in deriving valuable insight into the contract portfolio and preventing contract value leakages.